Oil inventory data has been one focus of oil price moves recently – along with financial crisis-related demand concerns and the US dollar, among others. According to the latest weekly EIA estimates published on Wednesday, inventories have built for the past seven weeks and last week reached an 18-year high of 370.6m. But American Petroleum Institute estimates, which are published every Tuesday, showed inventories fell by just over 1m barrels.
Until earlier this year, both organisations published their reports on the same day of the week. Today’s Schork Report speculates that the EIA statistics, coming from a government rather than an industry body, are taken more seriously, and some traders are trying to profit from the one day gap between the API and EIA figures.
But as they point out, trading on this difference is a difficult game. The mean average in the differences puts API’s figure at -.994m compared with the EIA’s. But actual results fall evenly on either side of that figure, meaning the API is just as likely to move above or below it.
Looking at the absolute difference between API and DOE figures in the box below, we see they follow a Gamma distribution. Think of the gamma as a symmetrical normal distribution with the left side (less than zero) chopped off.
This presents a better insight in to the previously calculated -0.994mmbls difference. We see that most of the values are clustered around this point, and the absolute mean is 2.3mbbls due to a few extreme outliers on the right hand side (9 mmbbls+) which occurred in 2007. This means that the absolute mean is deceptively high, and the real value is likely to be far smaller. The Gamma distribution is often used to model waiting times, say at a bus stop. We know the bus will usually arrive on schedule (give or take a few minutes), but on occasion there’s an accident and we’re waiting for hours. Similarly, we should expect the difference between API and DOE releases to be a minimal 1mmbls, and shouldn’t let a few extreme observations define the norm.
The bottom line is that the API figures have an absolute margin of error of just over 2.3 mmbls. That may seem small when we look at total inventories, which reach the hundreds of millions, but anyone hoping to exploit the weekly change, which averages 2.06 mmbbls for the DOE, could be wiped out with an alarming frequency.
API’s statistics are published today at 4.30pm EDT (9.30pm BST) and EIA data tomorrow at 10.30am (3.30pm BST). Platts’ survey of analysts shows a consensus forecast of a 1.8m barrel build and Reuters’ survey has 2.2m.
To read more, Schork’s oil outlook is usually published on the CNBC Guest Blog later in the day.
Updated: API wanted to point out that their surveys receive a response rate of more than 85 per cent, as do the EIA’s. They say their estimates are as good, or even better, than the EIA’s.